Effects of Varying Lengths and Diameters During Straw Phonation on an Excised Canine Model

Abstract

Objective/Background

Straw phonation, a well-established method of vocal exercise, started over one hundred years ago. However, some of the most basic questions about best practices remain including the ideal extension length and the ideal restriction (semi-occlusion) diameter. This study aims to help answer these questions by looking at both separately, with all other variables controlled.

Methods

Four (4) straws with lengths of 5, 25, 50 and 75 centimeters and four restrictions with diameters of 3, 9, 15 and 21 mm were tested, separately. They were attached to the end of a simulated vocal tract which was then affixed to an excised canine larynx. Airflow was increased until the larynges achieved self-sustained phonation then aerodynamic, acoustic, and electroglottographic data were collected.

Results

There was a significant decrease in the phonation threshold pressure using the 5 and 25 cm straw lengths and the 9 and 15 mm restrictions. The phonation threshold flow and contact quotient saw insignificant changes, except in the case of the 3 mm restriction.

Discussion

It is theorized that the increased inertive reactance helps to decrease phonation threshold pressure. By controlling for all other variables, the effects of the length or diameter of the straw can be analyzed in isolation. These results narrow the significant range of lengths and diameters used in straw phonation.

1. Introduction

It is estimated that almost a third of the population will be diagnosed with some sort of vocal disorder in their lifetime¹. This high prevalence creates a high economic cost resulting from lost wages and other factors². In addition to the economic cost, vocal disorders also create pain and discomfort for many people, making it harder to interact with others. Therefore, proper voice therapy is of upmost importance. One method for easing these problems, which is commonly used in Finland and is gaining prominence in the United States, uses commonly available things like drinking straws, stirring straws, and commercially available glass resonance tubes to achieve a semi-occluded vocal tract (SOVT)³. This version of semi-occlusion is commonly known as straw phonation therapy.

The theoretical framework for straw phonation was first published in 1904 and began to be popularized in the 1960s. During a straw phonation exercise a patient phonates through a straw, which can can help patients with vocal problems ranging from benign lesions, to patients with neurological diseases like Parkinson’s^4–6. It can help familiarize patients with the feeling of resonant voice, providing them with a goal to aim for when outside of vocal therapy. Straw phonation can also decrease the transglottal pressure and the phonation threshold pressure (PTP), which makes it easier to phonate^7–11. This decrease in PTP is theorized to be due to the increase in vocal tract inertive reactance^5,11–14. Reactance is a part of impedance and is the opposite of resistance in that it stores energy and makes phonation easier. It helps initiate vocal fold oscillation and aids their continued vibration. Reactance has a positive and a negative component, inertance and compliance. Inertance helps induce and continue phonation while compliance works against it^13–15.

In addition to providing these aerodynamic benefits, straw phonation can help in subjective measures of vocal improvement. Many participants report that immediately following an exercise with straws, phonation feels easier and more efficient or comfortable^9,16. The contact quotient (CQ), measure of the closure of the vocal folds, is also theorized to be impacted by SOVT exercises³. CQ has been shown to be a good indicator of vocal fold impact stress as it measures the duration of vocal fold closure during phonation¹⁷. Human studies looking at the effect straw phonation has on the CQ have found often contradictory things, some studies finding it decreases the CQ and others finding an increase or non-significant changes^3,18,19. This study is the first to look at CQ and straw phonation in a simulated excised model.

Some of the more basic questions about the practice of straw phonation remain unanswered. Researchers and clinicians tend to use straws that are already readily available, such as stirring straws, LaxVox ® tubes, glass resonance tubes, drinking straws and so on^3,4,14,20. While all these different tubes and straws do have varying lengths and diameters, they tend to be within a small range, usually about 11 cm to 30 cm long and about 2 mm to 9 mm in diameter^4,21. Because of this, there are currently very few straw phonation studies in the literature that focus on the diameter or the length in isolation. Those studies that do also tend to have a small scope or focus only on human subjects and are therefore limited in the data they collect ^{5–7,14,18,22}. Using excised canine larynges, we can have more control over the variables than testing on humans. The aim of this paper is to elucidate the range of lengths and diameters which produce the greatest improvement of aerodynamic variables, aiding clinicians and researchers in improving current SOVT therapy procedures. Using a wide range should aid these goals. The authors hypothesize that there will be a decrease in PTP as well as the phonation threshold flow (PTF) relative to the control on conditions which present a significant change to the vocal tract (e.g. a large extension or a small restriction).

2. Methods

29 larynges were excised from canines sacrificed for reasons unrelated to our research. These were donated from Covance Labs (Covance, Inc, Madison, WI). All procedures were performed in compliance with relevant laws and guidelines.. Canine larynges are known to be a good model for the human larynx due to their similarities in size and vocal fold histology, and they have been used to obtain much information about the function of the human larynx^23–25. They are also more inexpensive and readily available than human ones²⁴. The excision of the canine larynges follows the protocol outlined by Jiang and Titze²⁶. After excision, larynges were mounted on an apparatus designed to hold excised larynges, hereafter referred to as the excised bench, in a sound-attenuated three-walled room to maintain a constant temperature, humidity, and to isolate data collection from outside sound interference. The bench apparatus was designed to simulate the human respiratory system and is as specified in Jiang and Titze²⁶. Airflow was passed through two humidifiers (part: HC-150; Fisher Scientific, Hampton, NH), to warm and humidify the air before passing through a pseudolung and on to the larynx. A hose clamp secured the trachea to the pipe leading from the pseudolung. The length of the excised trachea was kept at 3 cm. The arytenoid cartilages were stabilized bilaterally using two three-pronged micrometer devices, allowing the vocal fold adduction to be completely controlled. Airflow was measured using an airflow meter (model FMA-1601 A; Omega Engineering Inc., Stamford, CT) and was manually controlled throughout the experiment by a needle valve. Subglottal pressure was measured through an Omega Pressure Transducer (part number MMG005V10W2MA0T1A4, Omega Engineering, Inc, Stamford, CT). Acoustic data was taken with an RTA-M microphone (part: 919; dbx by Harman Professional Solutions, Northridge, CA) connected to a Symetrix amplifier (Model 302; Symetrix Corporation, Mountlake Terrace, WA). The CQ of the larynx was recorded using a Laryngograph Electroglottograph (Laryngograph Ltd, Greater London, UK) and two custom four-pronged electrodes which were held bilaterally with secure arms on the posterior cricoarytenoid muscle. All data collection hardware was attached to a data acquisition board (model USB-6229; National Instruments Corporation, Austin, Texas). The input from the board was filtered and visualized through a custom LabVIEW (Nation Instruments Corporation, Austin, Texas) program which recorded all inputs. Figure 1 shows the excised bench setup.

Figure 1.

Schematic of the excised bench setup. The electroglottograph electrodes sat on the posterior cricoarytenoid muscle and could be retracted when not in use as shown in the figure. Figure 2 shows a more detailed schematic of the simulated vocal tract.

The simulated vocal tract was designed in SolidWorks (Dassault Systems, Waltham, MA) and made of PVC. The simulated vocal tract had three parts, the first being a laryngeal insert which allowed secure placement of the larynges, which was 30 cm long and had an inner diameter of 6 mm. The second part contained a conical transition (25 mm long) to a pharynx tube (9.5 cm long and 25 mm in diameter) which contained a small hole where supraglottal pressure was measured. It also contained an inlet and one-way valve which could provide airflow input. Since this study did not require a simulated oppositional supraglottal airflow, this valve remained closed. The third part was 5 cm long, 26.6 mm in diameter and had the insert into which the straws were placed. The larynges were affixed to the insert approximately 1 mm below the cuneiform cartilages using tissue glue (Loctite 401 Instant Adhesive, Part No. 40104; Henkel Corporation, Düsseldorf, Germany). The cuneiform cartilages were glued together to avoid air leakage. After the laryngeal insert, there is a conical transition to a larger tube where the supraglottal pressure was measured with a Honeywell ±1 cm H₂O pressure sensor (Honeywell International Inc, Morris Plains, NJ). For more detailed specifications of simulated vocal tract, see Conroy et al, 2014⁷. Figure 2 shows a schematic of the simulated vocal tract. The extensions were made of PVC pipe with an interior diameter of 20.5 mm and were attached at the outlet of the simulated vocal tract. The extensions had an interior diameter that was approximately the same as the interior diameter of the vocal tract in order to analyze the effect that extensions had without any semi-occlusion. The four extension lengths used were 5, 25, 50 and 75 centimeters. The semi-occlusions were also made with PVC, and they fit tightly in the end of the straw insert. The inner diameters of the semi-occlusion restriction inserts used were 3, 9, 15, and 21 mm. The restrictions included a very small extension (a ‘lip’) to aid in their removal from the simulated vocal tract, however, it was less than 4 mm and therefore had no significant effect on the outcomes. The lengths were chosen because they represented a wide range and included a range of commonly used lengths and a range of lengths which are rarely used in traditional exercises⁴. The restriction diameters were chosen in much the same way.

Figure 2.

Schematic of the simulated vocal tract as well as schematics for the length and restriction straw inserts. There is an inlet on the vocal tract where it is possible to attach an oppositional supraglottal airflow, although for this experiment this was not used. The inner diameter of the straw length was 20.5 mm and was variable for the straw restrictions.

Data collection was done in two parts. The variations in lengths were completed first and the variations in semi-occlusions were done later. The extension tests may have affected restriction testing, however, since we are looking at changes from a control condition, these effects will not affect our results. There were ten total conditions; one control which was followed by the experimental conditions; the four extensions or the four diameters; and finishing with a second control. The first control consisted only of the larynx and the simulated vocal tract as pictured in figure 2. The orders of the length or diameter conditions were randomized to reduce any confounding effects such as fatigue that could result from the order of conditions. The lengths or restrictions for these experimental conditions were placed in straw insert at the top of the simulated vocal tract. For the second control condition there was again no additional length or restriction on the vocal tract. The same 29 larynges were used for both the lengths and restrictions portions of the project and when not in use the larynges were kept frozen at −16°C in 0.9% saline. To start a trial the flow was slowly increased using the needle valve until the larynx reached phonation, at which point the flow was held at a constant level. For our purposes phonation was defined as when the harmonics and fundamental frequency were audibly stable. This was supported by a visual display of the spectrum. Figure 3 shows an example display of the acoustics and frequency. Flow was slowly increased until phonation was achieved, at this point the pressure and flow values were taken to be the PTP and PTF, respectively. The larynges were brought to phonation five times for five seconds during each condition and results were averaged across the five phonations. To minimize dehydration after each condition was tested, the folds were hydrated with 0.9% saline solution and were left to sit for five minutes to reduce fatigue. The second control condition was to ensure that data collection was consistent throughout the experiment and to examine if there was any effect of fatigue.

Figure 3.

Example of visual output for acoustic waveform (top) and the frequency spectrum (bottom) on the custom LabVIEW program used for data collection. This would be considered phonation and data would be collected.

2.1. Data Analysis

The custom LabVIEW program calculated the average results for each of the conditions. All ANOVA analysis was performed in SPSS v25(IBM Corporation, Armonk, New York). Overall repeated measures ANOVA was done with Greenhouse-Geisser correction. To assess significance between conditions, repeated measures ANOVA with Holm-Sidak correction was used. This compared all conditions of a variable pairwise. If normality assumptions for ANOVA were not met the data were transformed in SPSS using log₁₀ or squared transformation. Non-parametric Friedman signed Ranks Test and Dunnett’s method were used if normality assumptions were still not met. These nonparametric tests were done in SigmaPlot (Systat Software, Inc, San Jose, California). Data analysis focused on the PTP, PTF, CQ and supraglottal pressure. An alpha level of 0.05 was used to assign significance.

3. Results

We have reported the data from the 3 mm restriction; however, in the majority of larynges, no phonation was achieved using this small a restriction. This was due to the magnitude of the supraglottal pressure created, which lowered the transglottal pressure to the point where phonation could not be reached. Results for this restriction, therefore, represent a minority of larynges (n=12). Data were only collected if phonation was achieved.

3.1. Phonation Threshold Pressure

For straws varying in length, PTP was significantly decreased in an overall repeated measures ANOVA (p<0.001). The 5 and 25 cm straws were both significant when compared pairwise to the first control condition (p = 0.036 and p=0.043 respectively). For straws varying in semi-occlusion an overall repeated measures ANOVA with the Greenhouse-Geisser correction was significant (p=.026) and the PTP was significantly reduced in the 3, 9 and 15 mm conditions. The 15mm restriction showed the largest magnitude of change from the control (excluding the 3mm restriction). It was hypothesized that the 3 mm restriction would show the largest difference as many studies claim it to be the most impactful diameter in humans^15,27,28. Table 1 shows summary and test data for the PTP, Figure 3 shows a graph of the PTP means.

Table 1.

Summary statistics and analysis of the PTP.

CONDITION	EXTENSION MEANS (SD)	P-VALUE*	CONDITION	RESTRICTION MEANS (SD)	P-VALUE*
CONTROL	11.31 (3.17)	-	CONTROL	11.51 (2.51)	-
5 CM	10.02 (2.62)	.036†	3 MM (n=12)	8.70 (1.92)	.028†
25 CM	9.77 (2.55)	.043†	9 MM	10.65 (2.56)	.003†
50 CM	10.46 (2.77)	.286	15 MM	10.60 (2.61)	.002†
75 CM	10.54 (2.96)	.739	21 MM	11.11 (2.59)	.589
POST CONTROL	10.70 (2.97)	.849	POST CONTROL	11.23 (2.69)	.995

Table 1 Shows PTP values in cmH₂O. The left side shows results of the straw extensions and the right side shows results for the straw restrictions. The 3 mm restriction worked only in 12 larynges, which represents less than half of the total larynges (n=29).

^*P-values from a repeated measures ANOVA with Sidak correction for a condition versus a control

^†Significant at the 0.05 level

3.2. Phonation Threshold Flow

For lengths, the overall ANOVA with Greenhouse-Geisser correction was non-significant (p=.073). None of the pairwise comparisons were significant either. Data had to be transformed by log₁₀ to meet normality assumptions. For restrictions the overall ANOVA using Greenhouse-Geisser correction was significant (p=.028). However, the only condition to achieve significance pairwise was the 3 mm diameter.

The moderate difference between the controls of the two condition types is interesting, however since conditions are compared to the control of the same type, this difference did not affect the results. A small, nonsignificant difference was also found between the first control and the post control, most notably in the restriction conditions. This may be the result of fatigue of the vocal folds. However, since the conditions were performed in a randomized order this effect will not impact the results. Table 2 shows the summary and test data for the PTF.

Table 2.

Summary statistics and analysis of the PTF

CONDITION	EXTENSION MEANS (SD)	P-VALUE*	CONDITION	RESTRICTION MEANS (SD)	P-VALUE**
CONTROL	12.38 (4.00)	-	CONTROL	10.65 (3.47)	-
5 CM	10.89 (3.89)	.436	3 MM (n=12)	6.99 (4.06)	.005†
25 CM	10.83 (4.30)	.301	9 MM	10.22 (4.18)	.75
50 CM	11.06 (4.15)	.206	15 MM	9.55 (3.88)	.014†
75 CM	11.08 (3.80)	.627	21 MM	10.20 (3.90)	.769
POST CONTROL	11.76 (4.55)	.996	POST CONTROL	10.21 (4.17)	.747

Table 2 PTF in L/min. Table 2 shows the means and standard deviations of the phonation threshold flow in addition to the results of repeated measures ANOVA. The left half shows the results for the straw extensions and the right half shows results from the straw restrictions. The 3 mm restriction worked only in 12 larynges, which represents less than half of the total larynges (n=29).

^*p-value using a repeated measures ANOVA with Sidak correction for a condition versus control. Data were transformed using a log₁₀ transformation in order to meet the normality condition.

^†Significant at the .05 level

3.3. Supraglottal Pressure

For lengths, a Friedman test showed significant differences between the median values of the conditions (p=.008). Dunnett’s pairwise method showed that the longer two straws, 50 and 75 cm, saw small, but statistically significant changes from the control conditions, which was expected. Non-parametric tests were used because even using transformation the data were not normal. For restrictions there were significant differences found in the 3, 9, and 15 mm restrictions using a repeated measures ANOVA (p<.001).

3.4. Contact Quotient

For lengths the overall ANOVA test was significant using Greenhouse-Geisser correction (p=.001). However, none of the individual conditions were significant when compared pairwise to the control.

For restrictions the overall ANOVA test was significant using Greenhouse-Geisser correction (p<.001), however pairwise tests were only significant in the 3 mm condition.

4. Discussion

The purpose of this study was to identify a range of straw lengths and diameters that would induce the greatest improvement of parameters associated with phonatory ease, namely the PTP, PTF, CQ and supraglottal pressure. The literature has few investigations into just the straw lengths or diameters used in straw phonation exercises using an excised model. Additionally, if a study did look at differing lengths or restrictions, a smaller range was used and the two parameters were combined so their effects could not be separated ^7,14,29. The use of an excised model made it possible to record data that would be otherwise near impossible to collect in humans.

The failure of a 3 mm restriction to achieve phonation in a majority of trials was surprising. It is known to work clinically in humans and there are no studies in the literature which investigate this small a diameter with an excised model. It is often reported as the most clinically impactful straw diameter for humans due in part to the magnitude of increases in supraglottal pressure^15,27,28_. We theorize that the main reason that the 3 mm diameter did not work in our model was that the supraglottal pressure produced was too high for phonation to start. If the supraglottal pressure is too high it can make phonation effortful and can ‘kill’ phonation by decreasing the transglottal pressure too much^4,30.Furthermore, a static vocal tract model cannot react to these changes, whereas a human vocal tract may go through significant changes while phonating through a narrow straw³¹. This may explain why phonation through a 3 mm straw works well with humans but not in our excised model. The increase in the supraglottal pressure may explain the significant decrease in the CQ as well, as increased supraglottal pressure may abduct the vocal folds^7,21.

It was found that the PTP was significantly decreased in the smallest two lengths, 5 cm and 25 cm, which is the range in which most studies looking at straw phonation have focused^4,7,14,21. One reason for this may be the increased source-filter interaction which can lower the PTP as compared to normal voice. Theoretical studies have found that the vocal tract increasing in length can heighten this interaction and thereby increase the inertive reactance of the vocal tract, which is the base reason for the PTP decrease^15,20,27,30.

In the longest two straws, 50 and 75 cm, the only significantly changed parameter was the supraglottal pressure. The lack of significant changes to the PTP or PTF was unexpected according to Mills et al. as well as Story et al., because inertive reactance, which facilitates self-sustained phonation, should increase as the effective length of the vocal tract increases, theoretically decreasing the PTP^14,15. Mills et al. was the only other study found that controlled the diameter and looked only at the length of the straws¹⁴. Their longest and most impactful straw was 30 cm, not far away from this study’s most impactful one at 25 cm. Since that was the longest straw considered in a controlled study, it remained unknown if a longer straw would have a greater effect. Story et al. claimed that, using their calculations, the ideal length for a straw was 50 cm, a calculation which was later used in Gaskill, et al. which also failed to reach significance^3,15. The increase in supraglottal pressure further complicates this question, as the increase in supraglottal pressure is cited as one of the major benefits of straw phonation^7,12,27,30. Part of this apparent discrepancy may be due to the magnitude of the changes, because although there was a significant increase, the changes themselves were still quite small, with a difference in the median values of the control and the 75 cm straw of only .00365 cmH₂O. As shown from the restrictions, an increase in the supraglottal pressure that small does not cause an improvement in aerodynamic variables^29,32. There is also some disagreement about whether shorter tubes or longer tubes provide more of a benefit. Simberg; and Bele et al. have alternatively claimed that if the straw is too wide and/or too long, it can in fact make phonation effortful for the patient, a claim which is supported by this research^3,4,30,33.

During the length-only testing, it was theorized that the lack of change in the CQ was due to this same lack of semi-occlusion. Previous studies have found very mixed results with regards to the changes in the CQ, with some finding an increase¹⁹, some finding no significant changes^3,18 and some suggesting a decrease in the CQ²¹. Titze et al posited that a decrease in the CQ might be found in straw phonation due to the increased supraglottal pressure helping to slightly abduct the vocal folds^7,21. However, there was no significant change found using the restrictions either. Even quite significant changes to the supraglottal pressure in the 9 and 15 mm restrictions failed to bring significant differences in the CQ. The 3 mm restriction had a significant difference in CQ, however, as mentioned previously these results are inconsistent. Another possible reason for the lack of change in CQ could be due to the inherent differences when using a static simulated system versus a human vocal tract. During straw phonation exercises a human vocal tract will change significantly^5,9,31,34. Changing the adduction of the vocal folds could change the CQ, but this was something outside the scope of our study.

Hard-walled, static models of the vocal tract are relatively common in the literature as they allow for collection of data which are difficult to obtain in humans and allow for control of a much larger array of variables^35–37. However, there are differences which complicate generalization of data to human subjects. As previously mentioned, human vocal tracts tend to change shape and/or size significantly while performing straw phonation exercises and humans can change the level of adduction of the vocal folds. However, despite these differences, aerodynamic variables are not significantly different after excision³⁸. Therefore, our results should still be of interest to researchers working with human subjects. The question of straw phonation’s effect on the CQ remains and should continue to be studied.

5. Conclusion

This study was focused on the collection of aerodynamic, electroglottographic and acoustic data from excised canine larynges undergoing straw phonation, specifically differences in those parameters during straw phonation across different straw lengths and restrictions. Four lengths and four restrictions were evaluated against two controls, with a wider range than has been previously looked at in the literature. The only parameter that was significantly changed was the PTP with the 5 and 25 cm lengths and the 9 and 15 mm restrictions. The 3 mm restriction had a significant reduction in the PTP and PTF but was too unreliable. The PTP decrease is theorized to be due to the acoustic interaction and increased inertive reactance which can decrease the PTP. It is suggested that a combination of these lengths and diameters constitute the ideal dimension range for clinical applications, as a reduction in PTP indicates less effortful phonation. Future research should continue to look at the effect straw phonation has on the CQ, which is not solved by this paper. More research is needed to further elucidate the ideal length and restriction diameter within this range of significance.

Figure 4.

Graph of the PTP means for the straw lengths on the left and the restrictions on the right. Note that the sample size for the 3 mm restriction was n=12, whereas all other conditions had sample size n=29.

Table 3.

Summary statistics and analysis of the CQ

CONDITION	EXTENSION MEAN (SD)	P-VALUE*	CONDITION	RESTRICTION MEAN (SD)	P-VALUE*
CONTROL	.409 (.040)	-	CONTROL	.379 (.029)	-
5 CM	.395 (.038)	.923	3 MM (n=12)	.337 (.034)	.003†
25 CM	.405 (.030)	1.00	9 MM	.375 (.027)	1.00
50 CM	.399 (.025)	.905	15 MM	.384 (.031)	.839
75 CM	.396 (.036)	.597	21 MM	.382 (.033)	1.00
POST CONTROL	.392 (.038)	.244	POST CONTROL	.382 (.035)	.999

Table 3 CQ (dimensionless). Table 3 shows the means and standard deviations of the closure quotient (CQ) in addition to the results of repeated measures ANOVA. The left half shows the results for the straw extensions and the right half shows results from the straw restrictions. The 3 mm restriction worked only in 12 larynges, which represents less than half of the total larynges (n=29).

^*p-value of a repeated measures ANOVA of the conditions versus first control.

^†Significant at the .05 level